DIA Annual Meeting – Preparing to Share CDISC SDTM - Celerion
Transcript of DIA Annual Meeting – Preparing to Share CDISC SDTM - Celerion
Accurate: The SDTM Data are Complete and Consistent with the source Information
Consistent: The define.xml and blankcrf.pdf are consistent with the SDTM data
Conformant: The SDTM Data Conform to the Applicable CDISC Standards
Objective
CDISC data standards can help get needed new drugs to patients by streamlining the FDA submission process and strengthening scientific collaboration in drug discovery. For that potential to be reached, CDISC standards must be accurately and efficiently applied. Our objective was to generate a comprehensive review process that addressed potential translation errors when programmatically mapping source information to CDISC SDTM supported by review documentation that would meet the needs of downstream data consumers.
MethOds
Processes associated with programmatically mapping source information to CDISC SDTM were evaluated to see what types of translation errors were possible. Those potential errors were categorized, and review methods (including documentation suitable for all downstream users) were specified by category.
Results OveRview
Review of programmatic mapping processes for translation errors indicated that the primary sources of error are:
Documentation that is Not Consistent with the SDTM Data
Not Conforming with the Applicable CDISC SDTM Standards
Misrepresenting The Source Data
Given those primary error sources, data review needs to establish that the sdtM data are:
Preparing to Share CDISC SDTM Data: A Practical Evaluation of Potential Error Sources and Effective Review ProcessesSteve Kirby, JD, MS 1 and Nancy Wang, PhD 2
1 ViroPharma Incorporated. 2 Celerion
Accurate: the sdtM data are complete and consistent with the source information
Review MethodsPrimarily driven by manual review of information as collected and as mapped. That manual review can be programmatically supported in a variety of ways. More automation (and a simpler review process) is possible when source information is consistent with SDTM standards.
conformant: the sdtM data conform to the Applicable cdisc standards
Review MethodsPrimary done programmatically. Good validation code easily available (some free, some for fee). Recent FDA statements indicate that they are using Open CDISC.
consistent: the define.xml and blankcrf.pdf are consistent with the sdtM data
Review MethodsBoth programmatic checks (variable characteristics/code lists in define consistent with the data) and manual review (Is the define.xml origin consistent with the blankcrf.pdf and with the data) are needed.
cONclusiONs Generating and applying a complete and accurate set of review steps for each category of possible translation error when mapping source information to SDTM makes it so conforming, accurate data and documentation can be consistently delivered.
Those review steps must ensure the data accurately reflects the source information, conforms with all applicable CDISC standards, and has associated documentation (define.xml and blankcrf.pdf) that is consistent with the data. Where CDISC guidance supports more than one mapping or documentation choice, local rules are needed to clarify the choice to be made.
Preparing to share CDISC SDTM data is more than just preparing accurate, conformant SDTM domains and documentation; it is being able to show a wide audience why you are certain that the SDTM data provided are accurate, conformant and consistent. Proving that a comprehensive review was completed in a way that will satisfy a broad audience of downstream users makes it so those review steps do not need to be needlessly repeated.
Thought should be given to having review documentation available in a form that will serve the needs of multiple audiences. While each specific detail of the review process is important, many end users are most interested in and best served by a concise overview of the process. One effective approach is to have the final signature document give an overview of substantive areas checked with further detail incorporated by reference.
AuthOR cONtActThe authors welcome any comments and questions. Please feel free to contact:Steve Kirby, Manager Data Standards Implementation, ViroPharma: [email protected] Wang, Associate Director, Biostatistics, Celerion:[email protected]
disclOsuReAuthors of this presentation have the following to disclose concerning possible financial or personal relationships with commercial entities that may have a direct or indirect interest in the subject matter of this presentation:Steve Kirby: Nothing to discloseNancy Wang: Nothing to disclose
Process Guidelines and Review documentation Prove consistency with • Signaturedocumentthatreferencesmanualreviewof
source references (blankcrf.pdf annotations, define.xml origin references) completed [“Origin Audit”]
• Documentedprogrammaticverificationthatdefine.xmlreferences (outside of header information) are otherwise consistent with the data.
Note: Additional verification of blankcrf.pdf and define.xml are also needed to ensure that all applicable formatting standards are met. [See, e.g. MSG-SDTM].
tips • Beawareoflimitationsinprogrammaticreviewof
consistency between define.xml and SDTM data as any gaps (such as domain labels) need to be otherwise supported
• Buildingdefine.xmlcreation/reviewintothemappingprocessor creating define.xml based on data nicely supports consistency
• Supportingdefine.xmloriginassignmentandblankcrf.pdf annotation with local rules facilitates effective, efficient application and review
tips Programmatic program verification tends to be complete and accurate - but be aware of limitations that need to be filled with manual review or manual documentation• Therecanbeadisjointbetweenthedefinedtermslistsinthe
check program and in the specific controlled terms reference used for mapping as release schedules can vary
• CheckprogramsnaturallylagCDISCstandardupdates• Automatedconformancecheckscastawidenetwith
warnings (especially related to controlled terminology).The specific standards and verification code to be used can be effectively integrated into planning for mapping; consider whether a process is needed to support updating standards applied if planned study durations are very long.
Process Guidelines and Review documentation
Ensure that all validation issues are explained (and related to the specific code and standards used) in a format that can be included in submission package [typically located in Reviewer’s Guide].
tips Printing all unique combinations of related variables as collected and as mapped (useful combinations tend to be within categories) usefully highlights potential mapping issues.
SDATE_ORIG CMSTDAT1 FEB 2011 2011-02-01JAN 2012 2012-012010 12 JAN 2012 2012-01-12
• ReviewersinthisareadonothavetobeexpertsinSDTMifsupported by content that makes the relation between SDTM data and original information easy to follow
• HarmonizingcollectionwithCDISCcontrolledterminologyminimizes substantive review effort
• Reviewcanbeeffectivelybuiltintothemappingprocessandeffectively linked with mapping specifications
• ValidationofCDISCbasedreporttablesbasedoncomparison with tables programmed using source data offers efficient verification of the most scientifically critical content
Process Guidelines and Review documentation
Prove substantive validation was completed with signature document that references substantive checks completed
what to checkblankcrf.pdf• AreannotationsconsistentwithSDTMmapping• Areannotationsconsistentwithdefine.xmlorigin[matchwith
“CRF” page references, consistent otherwise]define.xml• Isdatadocumentationconsistentwithdata - Domain, Variable and Value attributes - Code list values• Isoriginaccurateandconsistentwithblankcrf.pdf - Origin references are consistent with SDTM mapping - CRF page references match blankcrf.pdf annotations - Accurate derivation comments support each derived variable
what to checkVerify data using published conformance criteria met. Criteria not met should be minimized, and deviations supported. Key information needed to support use of tools:• ExactversionofSDTMusedformapping•VersionofControlledTerminologyusedformapping•Whatversionsareappliedbycheckprogramused
what to checkCore Manual Check Categories Identified:• 1to1relations(includingexactmatch)• Not1to1,morecontextisneeded• Date/timeconversion• Missingdatareferences• Is“Assigned”and”Derived”contentcorrectandconsistent with any associated mapping specifications or documentation• Completeness–Arealldatarelevanttosubmissionincluded.
“Permissive” variables are key for this category.
The define.xml origin should be CRF if and only if :– There is a direct (exact or functionally equivalent), 1
to 1 match between data as collected and SDTM data.
– There is a direct (exact or functionally equivalent), 1 to 1 match between information pre-printed on the CRF and SDTM data.
Please see above for definition of “functionally equivalent”
CDISC Standards Applied SDTM Version 1.2 SDTM IG Version 3.1.2 Controlled Terms Version 22JUL2011
Conformance Code Used Program Open CDISC Validator Program Version 1.2.1 Program Inputs SDTM domains and define.xml
Conformance Findings
Type Message Notes Warning Value for MHBODSYS
not found in SOC controlled terminology codelist
MedDRA body systems as submitted are in mixed case as provided by the dictionary. The term list used for this check was in upper case.
www.viropharma.com